{"id":"b3f10f94-07bc-4bd3-a3ce-1ec832e99176","arxiv_id":"2412.14046","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":2,"one_line_summary":"Using 140 fb^-1 of 13 TeV pp collisions, ATLAS finds no excess and sets 95% CL upper limits of 0.2% to 2% on B(H to aa to gamma gamma tau tau) for m_a from 10 to 60 GeV.","lead":"ATLAS searched for exotic Higgs decays into two new light particles, where one decays into two photons and the other into two tau leptons. No signal was found, and the result sets new upper limits on the rate of this decay.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Background closure is the main residual risk, but it is controlled by the spurious-signal systematic; the verdict should stand unchanged.","rationale":"The reader identified the background modelling as the weakest assumption, and I agree: it is the most load-bearing element for the central null result and limits. However, the paper addresses this with a data-driven closure test, template variations, Gaussian-process smoothing, and a small spurious-signal systematic that is incorporated into the final uncertainty. The remaining risk is generic and not demonstrated by any evidence in the paper. The ggF-only assumption is a caveat but actually makes the limits conservative, so it does not threaten the validity of the quoted upper bounds. No internal inconsistency or critical flaw was found, so the ACCEPT verdict should stand unchanged.","tokens_in":48217,"tokens_out":5472,"duration_ms":50480,"concrete_test":"Refit the observed m_γγ distribution in the 10–60 GeV search window with an independent background model, such as a Bernstein polynomial of degree 4–6 or a kernel-smoothed empirical template from the inverted di-τ sideband, and recompute the 95% CL limits on B(H→aa→γγττ) for each m_a hypothesis. If any limit shifts outside the envelope implied by the quoted spurious-signal systematic (0.08σ_stat), the background closure assumption would need to be revisited.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The search's central claim relies on the background m_γγ distribution in the signal region being well described by the sigmoid-times-exponential analytic function (Section 5.2). The spurious-signal systematic is derived from templates built with inverted di-tau selections and smoothed with a Gaussian process whose length scale (1–2 GeV) is comparable to the signal resolution (0.2–1.0 GeV). If the true background deviates from this functional family in a way not captured by the template variations, the fitted signal yield and the quoted 95% CL limits would be biased. The quoted bias is small (<8% of the expected statistical uncertainty), and the closure tests are standard and thorough, so this is a residual risk rather than a demonstrated flaw. Still, it is the least secure link in the argument that the observed limits are unbiased and is therefore the most load-bearing assumption in the analysis.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a search for exotic Higgs boson decays H→aa→γγτhadτhad using 140 fb−1 of pp collisions at √s=13 TeV recorded by ATLAS in Run 2. The analysis targets a-boson masses between 10 and 60 GeV, reconstructing the diphoton system with a boosted selection and the hadronic tau pair with either resolved tau leptons or a dedicated boosted di-tau tagger. The background is estimated from data with an analytic sigmoid-times-exponential function, validated on control templates; signal shapes are taken from Monte Carlo simulation. No significant excess is found, and 95% CL upper limits on B(H→aa→γγττ) are set in the range 0.2%–2% observed and 0.5%–1% expected.","tokens_in":48419,"tokens_out":8352,"duration_ms":80042,"significance":"If the result holds, this is the first constraint on the γγττ final state in the H→aa search program and extends coverage to a-boson masses down to 10 GeV. The analysis uses the full Run 2 dataset, a data-driven background with a spurious-signal closure test, and a complete systematic accounting; the dominant 63% systematic on the boosted di-tau object at ma=10 GeV and the 27% theory uncertainty at 50 GeV are reported transparently. The stress-test concern about background closure is a residual modelling risk rather than a demonstrated flaw, because the spurious-signal systematic is evaluated on multiple template variations and found to be below 8% of the expected statistical uncertainty.","major_comments":[],"minor_comments":[{"comment":"The abstract describes the limits as 'model-independent', but Section 3 states that only the ggF Higgs production process is considered and other production modes are ignored; since the branching-ratio extraction assumes the ggF cross-section, I suggest clarifying that the limits are model-independent with respect to the a-boson decay model but assume SM Higgs production via ggF.","section":"Abstract and Section 3"},{"comment":"The background modelling uncertainty is quoted as '<0.08 σstat' and '0.16 events'; the text should define σstat as the expected statistical uncertainty of a fitted signal and explain how the spurious-signal envelope enters the likelihood as a nuisance parameter.","section":"Section 6 and Table 1"},{"comment":"The local p-values of 2.2σ at ma=39 GeV and 2.1σ at ma=48 GeV are reported without a trial-corrected global significance; since the search scans multiple mass hypotheses, a brief statement on the look-elsewhere effect would aid interpretation.","section":"Section 7 and Figure 5(a)"},{"comment":"The boosted di-tau scale factor extrapolation uncertainty of 50% is a major contributor to the low-mass systematic; the text should state whether this uncertainty is treated as fully correlated across mass points and how it is constrained in the fit.","section":"Section 4.1"},{"comment":"The bottom panel label 'Data-Fit' is not fully consistent with the caption text; please specify the units and the definition of the uncertainty band shown.","section":"Figure 4 caption"}],"recommendation":"accept","confidential_remarks":"No concerns about citation pattern or novelty; the manuscript is well within the journal's scope. The ggF-only production assumption is a minor limitation that should be clarified but does not affect the validity of the central result."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First things first: this is a solid, well-executed search, and the headline result—first limit on H->aa->gammagamma tau_had tau_had at 0.2% to 2% for ma between 10 and 60 GeV—holds up. The analysis uses the full 140 fb^-1 Run-2 dataset, a sensible event selection, and a data-driven background model with a spurious-signal systematic that is honestly quantified. The dedicated boosted di-tau tagger is a genuine technical step, and its 63% systematic at ma=10 GeV is large but not hidden; it is propagated properly. The 27% theory uncertainty at 50 GeV is also stated plainly.\n\nThe main soft spot is the background shape in m_gamma gamma. The signal region is fit to a sigmoid-times-exponential function, with templates built from inverted di-tau selections and smoothed using a Gaussian process whose length scale (1-2 GeV) is comparable to the signal resolution. If the true background deviates from this functional family in a way the template variations do not capture, the fitted signal yield and the 95% CL limits would be biased. That is the least secure link in the chain. But the paper quantifies the risk: the spurious signal is less than 8% of the expected statistical uncertainty, and several template variations are tested. So this is a residual risk, not a demonstrated flaw, and I do not think it invalidates the result.\n\nTwo smaller caveats. The limits are interpreted for ggF production only; other Higgs production modes are not included, which makes the interpretation somewhat model-dependent, but this is standard for such searches and clearly acknowledged. Also, the observed 2.2 sigma and 2.1 sigma excesses at 39 and 48 GeV are consistent with background fluctuations; the paper handles them correctly.\n\nBottom line: the math, data handling, and systematic accounting look sound. The citation pattern is appropriate, and the claims are correctly scoped. This is exactly the kind of paper that deserves a serious referee: I would accept it for publication, with at most minor requests like a spurious-signal scan as a function of ma as a cross-check. I would be happy to have this in the literature.","headline":"First combined gamma-gamma tau_had tau_had search with an honest background systematic; the residual shape risk is real but controlled, and the limits are what they say.","tokens_in":48929,"tokens_out":3040,"would_cite":true,"duration_ms":26874,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["13.85.Qk","14.80.Bn","14.80.Cp"],"model":"deepseek-v4-flash","headline":"This paper reports the first search for Higgs boson decays into two pseudoscalars where one decays to two photons and the other to two hadronically decaying tau leptons, finding no significant excess and setting 95% CL branching-ratio…","keywords":["Higgs boson exotic decays","pseudoscalar particles","diphoton final state","tau lepton pairs","boosted di-tau tagger","ATLAS","13 TeV proton-proton collisions","branching ratio upper limits"],"falsifier":"Look at a high-statistics control sample in the same mass range, for example events with same-charge di-tau pairs or inverted photon identification but otherwise identical selections; a localised bump in the $m_{\\gamma\\gamma}$ distribution in that sample at the same mass as the reported 39 or 48 GeV excesses, larger than the assigned spurious-signal systematic, would show the background model can create fake signals and would weaken the quoted limits.","tokens_in":48043,"feed_emoji":"⚛️","tokens_out":6113,"duration_ms":55545,"temperature":0.7,"pith_summary":"The paper tries to establish whether the 125 GeV Higgs boson can decay into a pair of new light pseudoscalar particles, $H\\to aa$, with one $a$ decaying to two photons and the other to a pair of hadronically decaying tau leptons. It is the first search for this $\\gamma\\gamma\\tau_{\\text{had}}\\tau_{\\text{had}}$ final state. Using 140 fb$^{-1}$ of 13 TeV proton-proton collisions, the analysis finds the observed diphoton mass spectrum consistent with the Standard Model background. It therefore converts the null result into model-independent upper limits on the branching ratio, ruling out $\\mathcal{B}(H\\to aa\\to\\gamma\\gamma\\tau\\tau)$ above 0.2% to 2% for $a$-boson masses of 10-60 GeV. A new reconstruction for collimated tau pairs extends sensitivity to low masses where the taus are boosted and overlapping.","feed_headline":"No excess in first search for Higgs to photons plus taus","feed_subtitle":"Using 140 fb^-1 of 13 TeV collisions, ATLAS sets 95% CL branching-ratio limits of 0.2-2% for a 10-60 GeV pseudoscalar.","key_machinery":"The carrying mechanism is a new dedicated reconstruction of collimated hadronic tau pairs: a large-radius jet is reclustered into subjets and a boosted decision tree scores the di-tau hypothesis using track and calorimeter substructure, with opposite-charge subjets required; this boosted di-tau tagger, combined with a resolved di-tau selection, recovers sensitivity at low $m_a$. The observable is the diphoton invariant mass, with signal modelled by a double-sided Crystal Ball function and background by a sigmoid-times-exponential function, smoothed by Gaussian-process regression, with a spurious-signal systematic derived from template variations with inverted tau selections.","core_discovery":"On its own terms, the central discovery claim is a null result with exclusion power: no significant excess of events is observed above the Standard Model background expectation in the diphoton invariant mass distribution, and the observed 95% CL upper limits on $\\mathcal{B}(H\\to aa\\to\\gamma\\gamma\\tau\\tau)$ range from 0.2% to 2% (expected 0.5% to 1%) for pseudoscalar masses between 10 and 60 GeV. The most significant local deviations, at $m_a=39$ and 48 GeV, have significances of 2.2$\\sigma$ and 2.1$\\sigma$, consistent with statistical fluctuations. This establishes the first constraint on the $\\gamma\\gamma\\tau_{\\text{had}}\\tau_{\\text{had}}$ final state and demonstrates that a dedicated boosted di-tau tagger can recover signal efficiency for $a$-boson masses below about 25 GeV.","pith_inferences":["Editorial inference: because the final state combines a photon-pair trigger and mass resolution with a mass-coupled tau-pair signature, the same analysis strategy could be combined with $H\\to aa\\to 4\\gamma$ and $\\tau\\tau bb$ channels to place joint model-independent bounds on the $H\\to aa$ coupling over a wider mass range.","Editorial inference: a natural next step is to apply the boosted di-tau tagger to higher-statistics control processes such as $Z\\to\\tau\\tau+$jets to calibrate the scale factor at the low $p_{\\text{T,vis}}$ regime directly rather than extrapolating with a 50% uncertainty.","Editorial inference: the local excesses at 39 and 48 GeV, while insignificant, define specific mass points where a future dataset could either confirm a signal or tighten limits, so a targeted follow-up scan in those bins would be informative."],"forward_implications":["The 95% CL upper limits of 0.2% to 2% on $\\mathcal{B}(H\\to aa\\to\\gamma\\gamma\\tau\\tau)$ are the first constraints on this final state and rule out pseudoscalar masses in 10-60 GeV with branching ratios above the quoted range.","Models of exotic Higgs decays, including two-Higgs-doublet and dark-matter scenarios, that predict $H\\to aa\\to\\gamma\\gamma\\tau\\tau$ rates above these limits are disfavoured.","The boosted di-tau tagger increases signal acceptance at low $a$-boson masses, where the two taus are collimated, and can be applied to other searches for boosted tau pairs.","The observed 2.2$\\sigma$ and 2.1$\\sigma$ excesses at 39 and 48 GeV do not reach the 5$\\sigma$ threshold and are interpreted as consistent with background."],"supporting_citations":[{"why":"Establishes the low-mass boosted diphoton search strategy, including the $p_{\\gamma\\gamma}^{\\text{T}}$ selection and the Gaussian-process smoothing used for the background templates.","marker":"[26]"},{"why":"Supplies the dedicated reconstruction and identification algorithm for collimated hadronic tau pairs (boosted di-tau tagger) that this search uses for the first time.","marker":"[69]"},{"why":"Provides the spurious-signal recommendation used to quantify the bias from the analytic background model choice.","marker":"[73]"},{"why":"Provides the two-dimensional sideband method used to decompose the background into $\\gamma\\gamma$, $\\gamma j$ and $jj$ components and estimate the diphoton purity.","marker":"[72]"},{"why":"Supplies the asymptotic profile-likelihood formulae used for p-values and for the expected limit calculation, validated with pseudo-experiments.","marker":"[76]"},{"why":"Supplies the CLs modified frequentist procedure used to set the 95% confidence-level upper limits.","marker":"[77]"},{"why":"Provides the gluon-fusion Higgs production cross section used to convert the fitted signal strength into a branching-ratio limit.","marker":"[75]"}],"fun_headline_variants":["No excess in ATLAS hunt for Higgs to photons plus taus","ATLAS sees nothing in Higgs to photons plus taus","ATLAS sets first limits on Higgs to photons plus taus","Higgs to photons and taus: no new physics from ATLAS","ATLAS constrains exotic Higgs decay to photons and taus"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result stands on the assumption that the background diphoton mass spectrum is smooth in the signal region and that its true shape is captured by the analytic function and the template variations used to estimate the spurious-signal systematic; if a real background structure deviates from this in a way those templates do not cover, the fitted signal and limits would be biased.","fun_headline_variants_meta":{"raw":{"variants":["No excess in ATLAS hunt for Higgs to photons plus taus","ATLAS sees nothing in Higgs to photons plus taus","ATLAS sets first limits on Higgs to photons plus taus","Higgs to photons and taus: no new physics from ATLAS","ATLAS constrains exotic Higgs decay to photons and taus"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00158,"raw_usage":{"total_tokens":6338,"prompt_tokens":1012,"completion_tokens":5326,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":628,"completion_tokens_details":{"reasoning_tokens":5238}},"tokens_in":628,"tokens_out":5326,"duration_ms":34871,"temperature":1.0,"reasoning_tokens":5238,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T12:31:36.163662+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Look at a high-statistics control sample in the same mass range, for example events with same-charge di-tau pairs or inverted photon identification but otherwise identical selections; a localised bump in the $m_{\\gamma\\gamma}$ distribution in that sample at the same mass as the reported 39 or 48 GeV excesses, larger than the assigned spurious-signal systematic, would show the background model can create fake signals and would weaken the quoted limits.","supporting_citations":[{"cited_title":"Reconstruction and identification of pairs of collimated $\\tau$-leptons decaying hadronically using $\\sqrt{s}=13$ TeV $pp$ collision data with the ATLAS detector","cited_arxiv_id":"2411.09357","evidence_quote":"Supplies the dedicated reconstruction and identification algorithm for collimated hadronic tau pairs (boosted di-tau tagger) that this search uses for the first time."},{"cited_title":"Measurement of isolated-photon pair production in pp collisions at sqrt(s) = 7 TeV with the ATLAS detector","cited_arxiv_id":"1211.1913","evidence_quote":"Provides the two-dimensional sideband method used to decompose the background into $\\gamma\\gamma$, $\\gamma j$ and $jj$ components and estimate the diphoton purity."}],"review_version":1}